PaperlayerView exact revision on arXiv

Source-linked AI summary

iFogSim: A Toolkit for Modeling and Simulation of Resource Management Techniques in Internet of Things, Edge and Fog Computing Environments

Harshit Gupta, Amir Vahid Dastjerdi, Soumya K. Ghosh, Rajkumar Buyya

arXiv:1606.02007v1cs.DC

TL;DR

IoT resource management needs repeatable evaluation for placing application components across heterogeneous edge-to-cloud devices while meeting real-time QoS requirements. iFogSim provides that simulation platform, with case studies demonstrating policy comparison and scalability evaluation.

  • Problem

    IoT analytics lacks repeatable evaluation of how application components should be scheduled across heterogeneous edge-to-cloud devices to meet real-time QoS requirements.

  • Method

    iFogSim simulates IoT, Fog, and cloud environments to evaluate resource-management policies for application placement, scheduling, and QoS-related metrics.

  • Results

    Two case studies demonstrate IoT modeling and resource-management policy comparison, while scalability is evaluated using memory consumption and simulation execution time.

  • Takeaways & Limitations

    iFogSim supports investigation and comparison of resource-management techniques for IoT, Fog, and Edge environments using latency and other QoS criteria.

  • Takeaways & Limitations

    Power-aware, priority-aware, and failure-aware Fog resource-management policies remain future research directions.

Abstract

from arXiv · show

Internet of Things (IoT) aims to bring every object (e.g. smart cameras, wearable, environmental sensors, home appliances, and vehicles) online, hence generating massive amounts of data that can overwhelm storage systems and data analytics applications. Cloud computing offers services at the infrastructure level that can scale to IoT storage and processing requirements. However, there are applications such as health monitoring and emergency response that require low latency, and delay caused by transferring data to the cloud and then back to the application can seriously impact their performances. To overcome this limitation, Fog computing paradigm has been proposed, where cloud services are extended to the edge of the network to decrease the latency and network congestion. To realize the full potential of Fog and IoT paradigms for real-time analytics, several challenges need to be addressed. The first and most critical problem is designing resource management techniques that determine which modules of analytics applications are pushed to each edge device to minimize the latency and maximize the throughput. To this end, we need a evaluation platform that enables the quantification of performance of resource management policies on an IoT or Fog computing infrastructure in a repeatable manner. In this paper we propose a simulator, called iFogSim, to model IoT and Fog environments and measure the impact of resource management techniques in terms of latency, network congestion, energy consumption, and cost. We describe two case studies to demonstrate modeling of an IoT environment and comparison of resource management policies. Moreover, scalability of the simulation toolkit in terms of RAM consumption and execution time is verified under different circumstances.

1. INTRODUCTION

The introduction motivates Fog computing for low-latency IoT analytics and identifies resource management as a central challenge. It presents iFogSim as a toolkit for evaluating Fog resource-management policies across latency, energy, network congestion, cost, and scalability.

  • IoT connects diverse devices, including medical devices, appliances, cameras, and sensors, to enhance services, infrastructure management, and disaster recovery.
  • Most current IoT processing transfers data to the cloud, despite limited support for real-time analytics and decision making.The passage attributes this limitation to analytics approaches designed for large data volumes rather than real-time processing and dispatching.
  • Distributed IoT applications traverse heterogeneous devices, making intermediate devices potential targets for computation offloading.These devices can provide computational and storage capabilities between data sources and cloud applications.
  • Resource-management policies are needed to support QoS, avoid energy wastage and resource fragmentation, and evaluate placement, migration, and consolidation techniques.The introduction calls for an evaluation environment for real-time analytics in Fog computing.
  • iFogSim models edge devices, cloud data centers, and network links to evaluate Fog policies by latency, energy consumption, network congestion, and operational costs.The paper also presents an IoT simulation recipe, two case studies comparing policies, and scalability evaluation using memory consumption and simulation execution time.

2. FOG COMPUTING — DEFINITION AND CONCEPTS

Fog computing extends cloud services to the network edge, coordinating cloud, edge, compute, networking, and storage resources. By processing sensor data near its source, it reduces cloud-bound traffic and propagation latency for real-time applications.

  • Definition: Fog computing is a distributed paradigm extending cloud services to the network edge while leveraging cloud, edge, and Fog infrastructure.It supports application components running across the cloud and devices between endpoints.
  • Definition: It manages and programs compute, networking, and storage services between data centers and end devices.
  • Benefits: Fog computing filters and analyzes sensor-generated data on edge devices, placing filtering operators close to the data source.This reduces the traffic sent to the cloud.
  • Benefits: Fog computing reduces propagation latency, particularly for mission-critical applications requiring real-time performance.

3. ARCHITECTURE

The architecture places hierarchical Fog nodes between geographically distributed IoT sensors and core cloud resources, enabling data processing across gateways and other Fog devices. It combines monitoring, resource management, and power monitoring with distributed application models for sensing, processing, actuation, and stream analytics.

  • Fog computing architecture: Fog nodes form a hierarchy between geographically distributed sensors and the cloud, while gateways connect sensors to the Internet and support processing and filtering.Sensors emit observed values upward through gateways; any network element capable of hosting application modules is a Fog Device.
  • Architecture services: Monitoring components track resource utilization, availability, application deployment, performance, and status across sensors, actuators, Fog devices, and network elements.They provide this information to other services as required.
  • Architecture services: Resource management coordinates placement and scheduling to satisfy application-level QoS constraints while minimizing resource wastage.Placement and Scheduler components use monitoring information to identify suitable candidates for hosting application modules.
  • Architecture services: Power monitoring evaluates application and resource-management impacts on energy consumption across heterogeneous Fog and IoT devices before production deployment.Fog environments contain many devices with heterogeneous power consumption, making energy management difficult.
  • Application models: Applications use the Distributed Data Flow model, representing processing modules and their data dependencies as a directed graph supporting Sense-Process-Actuate and Stream Processing models.Sense-Process-Actuate sends processed commands to actuators, whereas Stream Processing continuously analyzes sensor streams and stores mined information for large-scale, long-term analytics.

4. DESIGN AND IMPLEMENTATION

iFogSim extends CloudSim’s event-driven simulation to model Fog devices, IoT applications, tuple execution, and resource-management policies. Its design represents physical resources and application dataflows explicitly, enabling configurable placement, scheduling, monitoring, and topology construction.

  • Architecture: iFogSim builds on CloudSim’s event simulation, using message-passing events to coordinate Fog computing components.The main iFogSim classes define the toolkit’s architecture and physical-topology model.
  • Physical and application models: FogDevice models hardware, connectivity, and resource scheduling for Fog devices linked to other devices, sensors, and actuators.Its attributes include memory, processing capacity, storage, uplink bandwidth, and downlink bandwidth.
  • Physical and application models: Applications are directed acyclic graphs whose modules process tuples, edges encode data dependencies, and loops specify end-to-end latency measurements.AppModule uses selectivity models to determine output tuples, while AppEdge supports periodic and event-based communication.
  • Execution flow: Tuple execution begins with sensor emission to a connected gateway, where incoming tuples are processed locally or routed immediately to another Fog device.The gateway invokes processTupleArrival() when a tuple arrives.
  • Resource management: Resource management separates application placement from host-level scheduling, with placement policies targeting latency, network usage, operational cost, or energy consumption.ModulePlacement is the abstract policy class for integrating new placement strategies.
  • Simulation workflow: Users can construct physical topologies through a GUI or programmatically, configure sensor transmit rates and tuple resource requirements, and define DAG-based placement and scheduling policies.Topologies can be saved and reloaded in JSON format.

5. APPLICATION CASE STUDIES

The section presents two iFogSim case studies: a latency-sensitive EEG game and intelligent surveillance using distributed camera networks. Together, they illustrate Fog computing application models and the need for low-latency, decentralized processing.

  • EEG Tractor Beam Game: The EEG Tractor Beam Game processes wireless-headset EEG signals in real time on smartphones and calculates each user’s brain state.Players exert force on a target according to concentration estimated from EEG α, β, and θ bands.
  • EEG Tractor Beam Game: The EEG application exemplifies Fog computing because smartphone hosting supports real-time processing, whereas cloud deployment enables global coverage.The case study models Client, Concentration Calculator, and Coordinator modules, along with their data dependencies and control loop.
  • EEG Tractor Beam Game: The EEG case study uses a physical topology with 4 Fog devices and two EEG headset types emitting tuples with different properties.The topology is modeled using FogDevice, Sensor, PhysicalTopology, and Actuator classes.
  • Intelligent Surveillance: Decentralized video analysis is preferred because centralized processing would incur high latency and consume available bandwidth when handling camera-generated data.The application comprises Motion Detector, Object Detector, Object Tracker, PTZ Control, and User Interface modules.
  • Intelligent Surveillance: The Intelligent Surveillance system coordinates cameras with different fields of view by tuning PTZ parameters and alerting users to irregular events.Smart cameras detect motion, forward video streams for object processing and tracking, and continuously adjust PTZ parameters for coverage.

6. PERFORMANCE EVALUATION

The performance evaluation simulates Fog computing environments for application case studies and assesses placement strategies and iFogSim scalability. It measures latency, network usage, energy consumption, RAM usage, and execution time across different scenarios.

  • The evaluation simulates a Fog computing environment for the application case studies.
  • It compares cloud-only and edge-ward placement strategies using latency, network usage, and energy consumption for each case study.
  • It evaluates iFogSim scalability by measuring RAM usage and execution time across different simulation scenarios.

6.1. Evaluation of Case Study 1 — A Latency-sensitive Online Game

The three-hour EEG Tractor simulation evaluates how workloads and placement strategies affect network usage, latency, and energy consumption. Fog-based deployment reduces network burden and cloud energy consumption, while slightly increasing edge-device energy use.

  • 6.1. Evaluation of Case Study 1 — A Latency-sensitive Online Game: The latency-critical control loop transforms the user’s brain state into the smartphone game state, requiring real-time communication and efficient classification processing.Lag in this loop severely harms user experience.
  • 6.1. Evaluation of Case Study 1 — A Latency-sensitive Online Game: Increasing device counts significantly raises network usage when only cloud resources are used, whereas including Fog devices considerably decreases network usage.The simulation collected iFogSim metrics over three hours while varying workloads and placement strategies.
  • 6.1. Evaluation of Case Study 1 — A Latency-sensitive Online Game: Fog-based deployment helps avoid uncontrolled network-usage growth, congestion, and further degradation of application performance.The result is interpreted as evidence of scalability for Fog-based applications.
  • 6.1. Evaluation of Case Study 1 — A Latency-sensitive Online Game: Using Fog devices with an Edge-ward placement strategy reduces cloud-data-center energy consumption while slightly increasing energy consumption at edge devices.The energy comparison covers different device classes in the simulation.

6.2. Evaluation of Case Study 2 — Intelligent Surveillance through Distributed Camera Networks

The Intelligent Surveillance case study evaluates cloud-only and Edge-ward placement across configurations with 1–16 surveilled areas, each monitored by four smart cameras. Edge-ward placement reduces control-loop latency and network usage by placing critical, data-intensive modules near the network edge, while camera energy consumption rises with surveillance scale.

  • Infrastructure and placement: The simulated topology connects four smart cameras per surveilled area to an area gateway, with cameras at the network edge and the cloud data center at the apex.Live video streams are modeled as tuples for motion detection, while PTZ control is represented by an actuator.
  • Latency: Cloud-only placement makes cloud data centers execution bottlenecks that notably increase average sensing-actuation control-loop latency.Figure 14 reports the average processing latency of the sensing-actuation control loop.
  • Latency: Edge-ward placement maintains low control-loop latency by placing modules critical to the control loop close to the network edge.This placement strategy is compared with cloud-only deployment in the case study.
  • Network usage: As the number of connected devices increases, cloud-only deployment produces significantly higher network usage than edge-ward deployment.Fog execution routes most data-intensive communication through low-latency links and reduces data sent to the centralized cloud.
  • Energy consumption: As surveilled areas increase, camera energy consumption also increases because motion detection in captured video frames drains substantial power.Energy consumption is compared between Fog-device deployment and deployment only on cloud data centers.

6.3. iFogSim Execution Footprint Analysis

iFogSim’s execution footprint was evaluated using RAM usage and simulation time across topology sizes and input workloads. The results indicate minimal memory overhead and near-linear execution-time growth, with simulations remaining acceptable at 25 seconds even when many gateways are added.

  • Execution Footprint Metrics: Scalability was assessed by measuring RAM usage and simulation execution time across simulation sizes and input workloads.The evaluation used the first case-study configuration, with Headset-B emitting tuples at twice Headset-A’s rate.
  • RAM Usage: Heap allocations did not increase considerably as workload and physical topology size increased.Massif heap profiling measured allocations while sensors increased from 4 to 64 and gateways from 1 to 16.
  • Execution Time: 25 seconds remained an acceptable simulation time even after adding a considerable number of gateways.Execution time increased with device count and transmission rate, but the increase was almost linear.

7. RELATED WORK

Prior work includes large-scale IoT testbeds and simulators focused on wireless networking, communication protocols, or service-based applications. Unlike these tools and cloud-only simulators, iFogSim models Fog together with IoT and cloud to evaluate resource-management policies end to end.

  • IoT testbeds: FIT IoT-LAB provides thousands of wireless nodes across six sites in France for evaluating ideas in a large-scale wireless IoT environment.Its services include remote access to sensors and gateways through APIs for firmware and application development.
  • IoT testbeds: SmartSantander offers a city-scale smart-city facility with many IoT devices deployed across urban locations and a three-tier architecture.The project belongs to the European Commission’s Future Internet Research and Experimentation initiative.
  • IoT simulators: WSNet models wireless-network and environmental phenomena, while SimpleIoTSimulator targets IoT communication protocols and product-quality improvement.SimpleIoTSimulator supports CoAP and MQTT, whereas the proposed simulators focus on application design and analysis.
  • Cloud simulators: CloudSim models cloud components such as virtual machines, data centers, and users, but does not model IoT devices or stream-processing applications.The passage also identifies GloudSim, DCSim, and GroudSim as cloud-environment simulators with the same limitation.
  • iFogSim’s distinction: iFogSim is designed to model Fog environments together with IoT and cloud, enabling comprehensive end-to-end evaluation of resource-management policies for real-time stream processing.Simulation is presented as a cost-effective, repeatable, and controllable step before real experimentation.

8. CONCLUSIONS AND FUTURE DIRECTIONS

The paper introduces iFogSim for modeling and simulating IoT, Fog, and Edge computing environments, which use edge devices near sensors to reduce latency and network congestion. Future work targets enhanced capabilities and resource-management strategies, including power-aware policies that migrate operators according to device battery life.

  • Conclusions: iFogSim models and simulates IoT, Fog, and Edge computing environments using edge devices closer to sensors to reduce latency and network congestion.The approach supplements cloud processing with devices that have more processing power than end devices.
  • Future directions: Future directions include enhancing iFogSim capabilities and developing resource-management strategies for IoT environments.The paper presents these as directions for extending the toolkit and its resource-management scope.
  • Future directions: Power-aware policies should dynamically migrate operators based on Fog-device battery life, addressing which operator to migrate, when to migrate, and where to migrate.The goal is to obtain additional battery life for Fog devices.
  • Future directions: Priority-aware resource-management strategies for multi-tenant environments are identified as another future direction.The supplied passage names this direction but does not provide further details.
Loading 1606.02007v1…